Carbon transition risk increasingly affects the financing conditions of firms in high-energy-consuming industries, yet debt-side evidence remains fragmented and often mixes policy inference with predictive monitoring. This study examines how carbon-related transition signals are associated with green debt financing spreads for Chinese A-share listed firms in high-energy-consuming industries during 2010–2024. The dependent variable is the green debt financing spread measured in basis points (Spread_bps). To preserve evidentiary boundaries, the study adopts a dual-track design: panel fixed-effects models, an event-study analysis around the 2015 Paris Agreement, and a segmented difference-in-differences design around China’s national emissions trading system in 2021 are used for policy-timing evidence, while random forest, extreme gradient boosting (XGBoost), and neural network models are used for out-of-sample prediction, ablation tests, Shapley additive explanations interpretation, and conditional scenario simulation. The econometric results show that higher carbon intensity is associated with wider green debt spreads, whereas stronger carbon-pricing exposure and greater green patent output are associated with narrower spreads. Pricing sensitivity to carbon-related factors increases after major policy milestones, and green innovation partly mitigates post-emission-trading-system financing pressure on high-carbon issuers. The predictive results show that XGBoost performs best and that removing carbon variables reduces forecasting accuracy, confirming the incremental monitoring value of carbon information. These findings suggest that green debt pricing can complement environmental regulation by translating carbon intensity, carbon-market exposure, and transition capacity into financing signals useful for credit analysts, issuers, investors, and policymakers.